What problem does it solve? Teams accumulate operational knowledge in README, RUNBOOK, AGENTS, and DECISIONS files, but that knowledge stays passive text instead of becoming reusable agent workflows. This Skill inspects a repository's documentation and current GitHub Issues/PRs, separates live operational recipes from stale or speculative material, and converts approved recipes into project-local agent skills under .agents/skills/. ## Core Features & Use Cases - Evidence-based discovery: Reads docs in authority order (AGENTS, RUNBOOK, DECISIONS, README, then GitHub Issues/PRs) and detects stale, contradictory, or speculative recipes before proposing anything. - Approval-gated generation: Produces a structured proposal of recommended, optional, unclear, and rejected candidates, then waits for explicit approval before writing any skill. - Hardened skill authoring: Generates one kebab-case skill per approved recipe under .agents/skills/<skill-name>/, adding approval gates and safety boundaries for dangerous procedures like deployments or migrations. - Use Case: Point it at a repository with a mature RUNBOOK and DECISIONS file; it proposes skills such as deploy-service or rotate-credentials, and after your approval writes validated project-local skills with activation criteria, verification steps, and source references. ## Quick Start Ask the agent to mine this repository's documentation and propose which project-local skills should exist under .agents/skills/.
